Product design question

Build a mobile app that takes a picture of a resume and recommends jobs based on that.

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Product design combining a computer vision feature with a recommendation engine, checking you address both technical feasibility and user trust.

Clarify the target user, likely job seekers wanting a faster alternative to manual resume entry, and confirm whether this integrates with an existing platform like LinkedIn or Glassdoor.
List pain points: manual resume upload and parsing friction, and generic job recommendations that don't reflect a candidate's actual experience nuances.
Describe the core mechanism: OCR and text extraction from the photo, parsing into structured resume data, then matching against a job recommendation model.
Address accuracy risk explicitly, since OCR errors could misrepresent a candidate's skills, requiring an easy review and correction step before recommendations are finalized.
Prioritize the review and correction step as essential, not optional, since trust in the recommendations depends on the underlying data being accurate.
Define success as recommendation relevance, measured by application rate on suggested jobs, and photo-to-profile completion rate, confirming scope on multi-page or multi-format resume support.

Explicitly addresses OCR accuracy risk and mandates a user review step before finalizing parsed data, since bad extraction would produce untrustworthy recommendations.
Describes the pipeline clearly, OCR extraction, structured parsing, then matching, showing the technical flow rather than treating it as a black box.
Names a concrete success metric, application rate on recommended jobs from photo-parsed profiles compared to manually entered ones, to validate the feature actually works as well as manual entry.
Considers format edge cases, like handwritten or non-standard resume layouts, as a known limitation worth flagging.

Ignoring OCR accuracy risk and assuming extracted data is automatically trustworthy.
Not including a user review or correction step before generating recommendations.
Treating the technical pipeline as a black box with no explanation of how parsing to recommendation actually works.

How would you handle a resume photo with poor image quality or unusual formatting?
How would you measure whether photo-based profiles produce recommendations as good as manually entered ones?